TWI361601B - Calibration of downlink and uplink channel responses in a wireless mimo communication system - Google Patents

Description

1361601 IX. INSTRUCTIONS: [Technical field to which the invention pertains] 2. The present invention relates to general §, and more particularly to calibration for downlink in a -V wireless multiple input multiple output (MIMO) communication system Link and uplink and downlink channel response techniques. ν [Prior Art] A system uses multiple (Ντ) transmit antennas and multiple (Nr) receive antennas for data transmission. A MIM channel composed of the Ντ transmit antennas and Nr receive antennas can be decomposed into Ns spatial channels, where Ns $ min {NT, NR}. These Ns spatial channels can be used to transmit data in parallel to obtain a higher overall throughput or to achieve higher reliability in a redundant manner. In order to achieve high performance, it is usually necessary to know the response of the Asahi elbow channel. For example, performing a spatial process on a downlink transmission to a user terminal' access point may require knowledge of the response of the downlink channel. In a conventional channel estimation technique, an access point transmits a φ-pilot on the downlink, and then the user terminal estimates a downlink channel response based on the pilot and estimates the downlink channel response. The value is sent back to the access point. Such channel estimation techniques use uplink resources and further cause a delay in sending back channel response estimates, both of which are not what we want. The TDD system is used for both the downlink and the uplink - a single frequency I band' where a portion of the downlink key is allocated, and the rest of the time is allocated for the uplink. For TDD systems, the downlink and uplink channel responses can be assumed to be reciprocal. In other words, if £[represents the channel response matrix from antenna array A to antenna array B, then the reciprocal channel means from array B to array 99590.doc

Shown as sT ’ which represents the transposition of the court. The channel response for the t-links (e.g., the downlink) for the -reciprocal channel' can be estimated from the pilot received via the other link (e.g., the uplink). > Access Point and User Terminal: The transmitting and receiving downlink transmissions are respectively performed using the transmit chain and the receive key, thereby observing the "effective" downlink channel response, the "active" downlink channel The response includes the response of the transmit chain at the access point and the receive chain at the user terminal. Accordingly, an uplink transmission will observe a valid uplink channel response that includes the transmit key at the user terminal and the response at the access point (4). The response of the transmit chain and the receive chain at the access point is different from the response of the transmit chain and the receive chain at the user terminal. Therefore, effective downlink and uplink channel responses are usually not mutually exclusive. If the obtained channel response estimate for one link is used to perform spatial processing on another link, any difference in the response of the access point to the transmit/receive chain at the user terminal will be reflected as an error, if Without determining and compensating for such errors, it may degrade performance. Therefore, there is a need in the art to provide techniques for calibrating the downlink and uplink channel responses in a TDD® system. SUMMARY OF THE INVENTION This document describes techniques for calibrating the downlink and uplink channel responses to compensate for differences in the transmit chain and receive chain responses at the access point and the user terminal. After calibration, the The channel response estimate for one of the links is used as the channel response estimate for the other link. This simplifies channel estimation and spatial processing. 99590.doc 1361601 This calibration can be divided into two parts—initial calibration and subsequent calibration. In the initial calibration, the access point and the user terminal transmit MIM〇 pilots on the downlink and uplink, respectively (as described below). These MIM〇 pilots are used to derive “effective” downlink and uplink channel response estimates, and *· iup, these “effective” downlink and uplink channel response estimates. Idn and &Up include the response of the applicable transmit/receive chain. The channel estimates and Eup are used to derive the correction matrix I and I, and then the correction matrix I and I are used by the access point and the user terminal to compensate for the response of the transmit/receive key, respectively, as described below. In subsequent calibrations, one of the entities (e.g., access points) transmits a pilot and a pilot reference signal (as described below). Another entity (e.g., user terminal) (1) derives an "actually received" transmit matrix 根据 based on the guided reference signal, and (2) based on the ΜΙΜ〇 pilot and calibration error matrix 仏ρ and redemption Deriving a "hypothetical" emission matrix, the matrix ^^ and 仏^ respectively contain conjecture values or estimates for the error in the correction matrix and the box (1). The difference between the φ 乞 matrix and the 咖 coffee represents the accuracy of the estimate of the error in the correction matrix. The matrix can be adjusted according to an adaptive procedure to minimize the error between 乞 and Yhyp. Several adaptive procedures for the iterative matrix s and the t are described below. Then, the correction matrix boxes ρ and L are updated by calibrating the error matrices Q_ap and Q_ut, respectively. / Various aspects and embodiments of the invention are set forth in more detail below. [Embodiment] The term "exemplary" is used herein to mean "serving as an example, instance or example." Any embodiment described herein as "exemplary" is not necessarily considered to be 99590.doc. Good or beneficial. • The calibration techniques described in this paper can be used for both single-carrier and multi-carrier TDD systems. • For the sake of beta, the techniques will be described below for a single-carrier TDD system. 1 shows a block diagram of an access point i J 一 of a TDD system 1 and a transmit and receive portion at a user terminal 150. For the downlink, at access point 110, the transmitted symbols (represented by a vector )) are transmitted from the Nap antenna 110, via a ΜΙΜ〇-transmit chain 114, via a wireless channel that responds to Η. At user terminal 150, the Nap downlink signals are received by Nut antenna 152 and processed by a receive chain 154 to obtain the received payout number (represented by -vector bn). The processing performed by the transmit chain 14 typically includes digital-to-analog conversion, amplification, filtering, upconversion, and the like. The processing performed by receive chain 154 typically includes downconversion, amplification, filtering, analog to digital conversion, and the like. For the uplink, at the user terminal 15, the transmitted symbols (represented by φ-vector Lp) are borrowed from the Nui antenna 152 after being processed by a transmit chain 164.

Launched by the channel. At access terminal 110, the Nut uplinks "is 5 are received by Nap antenna 116 and processed by a receive key 124 to obtain received symbols (represented by a vector bp) C for the downlink In terms of roads, the receiving vector at the user terminal can be expressed as

Idn =Eu.HTap2(l„ 5 Equation (1) where the branch has the Nap transmit symbols transmitted from the Nap access point antennas 99590.doc l36l6〇l

Vector; Ldn is a vector having Nut received symbols obtained through Nut user terminal antennas; EaP is a Nap X Nap order diagonal matrix having Nap complex gains corresponding to the access point transmit chain, wherein each - the access point antennas all correspond to a complex gain;

Eut is a Nut X Nut diagonal matrix with Nut complex gains corresponding to the receiving chain of the user terminal, wherein each user terminal antenna corresponds to a complex gain; and the Down X Nap of the SL system downlink Order channel response matrix. The response of the transmit/receive chain and the ΜΙΜΟ channel is usually a function of frequency. For the sake of simplicity, it is assumed to be a flat fading channel with a flat frequency response. For the uplink, the receive vector at the access point can be expressed as:

Tup = EapHrTmXup, where 2Lup has a vector from Nut numbers; Equation (2) Transmitter transmitted by the user's terminal antenna

The Hup system has a vector mNut order diagonal matrix of received symbols via Nap access points, which has a team corresponding to the user's transmit chain (a complex gain, and a complex gain; a mother user terminal) The antennas are all connected to a Napx Nap-order diagonal matrix, which has a chain of Nap complex gains, where each #点点= should be in the access point. The mother-access point antenna corresponds to a 99590.doc •10 - 1361601 Gain; and Ting 7" is the uplink Nap X Nut order channel response matrix. According to equations (1) and (7), effective "downlink and uplink channel response" including applicable transmit and receive key responses Heart and &up can be expressed as.

Sdn=EiaS3:*p and H„p=EapHTI:m. Combine the two equations in equation group (3) to get the following formula:

Equation (3) ' ^ ' Equation (4) where '=ϊ>ρ and Kut=r> ut. I. The ap-access point-NapXNaf-order diagonal matrix, which responds to the transmit chain by the receive chain response Eap The ratio of Xap is obtained to obtain 'the ratio is obtained from element to element. Similarly, a ratio of the NutXNut order diagonal matrix of the ILu user terminal, which is the response of the receiving chain by the receiving key response Eut Equation (4) can also be expressed as: Equation (5) where ilcup represents the calibrated channel response of the uplink; and SLcdn represents the calibrated channel response of the downlink. Matrix Kap and Etu are included for compensation access The value of the difference between the transmit/receive chain at the point and the user terminal. As shown in equation (5), applying the diagonal matrix base and &^ to the effective downlink and uplink channel responses will cause The calibrated channel response of one link can be expressed by the calibrated channel response of the other link. 99590.doc 8 1361601 An initial calibration can be performed to determine the matrix Hap and Κ〆 normally, the true channel response and the transmit/receive chain The response is not my It is known and cannot be accurately determined. - Conversely, the effective downlink channel response can be estimated based on the MIM pilots transmitted on the downlink and uplink* links respectively. The channel channel responds to the ILP. A MIM〇 pilot system is a pilot consisting of Ντ pilot transmissions transmitted from a self-transmitting antenna, wherein pilot transmissions transmitted from each transmitting antenna are recognized by the receiving entity. The use of φ, for example, for pilot transmission from each transmit antenna is achieved using a different orthogonal sequence. Then, as described below, the estimated downlink and channel uplink response values can be based on the estimated feed center and uplink. The channel response estimate is good p to derive the matrix Kap and U which are called the correction matrix ap and the estimate of the one.

Kap and Eu contain correction factors that compensate for the difference between the access point and the transmit/receive chain at the user terminal. 2 shows that the correction matrix Kap is used at the access point 11 and the user terminal 15 and on the downlink, the transmission vector Ln φ is first multiplied by the correction matrix, p, by a unit i12. The subsequent implementation by the transmit chain π 4 and the receive chain 154 is illustrated in FIG. Similarly, on the uplink, the transmit vector Lp is first multiplied by a unit 162 by the correction matrix & The subsequent processing by transmit chain 164 and a receive chain 124 is also illustrated in FIG. In the ΜΙΜΟ system, the material can be transmitted on the Ns eigenmodes of a MIM 〇 channel. The eigenmodes can be considered as orthogonal spatial channels of the chirp channel. The channel response matrix H r can be diagonalized to obtain the Ns eigenmodes of the μιμ〇 channel. Such diagonalization can be achieved by performing singular value decomposition on the singularity or on the correlation matrix of Η (which is the conjugate 99590.doc * 12- 160 1601 transpose of η) . Table 1 shows the effective channel response and calibrated channel response for the downlink and uplink and the singular value decomposition for the calibrated downlink and uplink channel response matrices. Table 1 - Singular Value Decomposition

Singular value decomposition for real channel response =Y; tZrU:p uplink

Hcup_ Hupl£u

Hcup=uapiv^ Singular value decomposition of the estimated channel response Unnormalized emission matrix Η^=ν;ΣΓϋ:

Uapt = HcupVu In Table 1, ilap is a NapXNap order unit matrix consisting of the left eigenvector of ilcup, which is a NapXNut order diagonal matrix composed of singular values of &cup, and Xut is one by |£eup The right eigenvectors form a matrix of order units, and the "*" table is not conjugated. The unit matrix cluster is characterized by the characteristic ^'s towel lookup, good matrix. The matrices 匕 and 仏 are also matrices composed of the left eigenvector and the right eigenvector, respectively. Matrix 罝, 死, and different forms of the matrix. For the sake of brevity, the matrices Hap and Xut mentioned in the following description may also relate to other forms thereof. The matrix 仏ρ and its also known as the transmit matrix are used by the access point and the user terminal for spatial processing and are thus marked by their subscripts.

Gilbert Strang is entitled "Linear Algebra and Its Applications (B (6) Algebra and Its AppIicati〇ns)" (Second Edition, —c 99590.doc -13- 8 1361601

The singular value decomposition is described in more detail in Press ’ 1980 (1980), which is incorporated herein by reference. In the actual system, the matrix can not be obtained... and 艮up. Instead, the user terminal can estimate the calibrated downlink channel response based on a MIM 〇 pilot transmitted by the access point. The user terminal can then perform singular value decomposition on the verified downlink channel response estimate pool to obtain a diagonal matrix! And - a matrix consisting of the left eigenvectors of the ground, and the hat symbol ("Λ") above each matrix indicates the actual matrix: the estimated value. Similarly, the 'access point' can estimate the calibrated uplink channel response based on a boost pilot transmitted by the user terminal. Then, the access = can perform a singular value solution on the calibrated uplink key channel response estimate ttcup to obtain a pair of angular matrices | and a matrix ϋ ρ consisting of the left eigenvector of the cup. Since it is a reciprocal channel and is calibrated, it is only necessary to perform a singular value decomposition by the user terminal or t-point to obtain the matrix £. For the π m # yoke scheme, the user terminal obtains the calibrated lower wire ^. The D-key channel response estimate is good, and the decomposition is performed on the Ting coffee, using the matrix iL for the *M by -cdn Inter-processing, the spurt pushes the matrix 〇ap back to save ''' with a homing reference signal. 3. A ^ takes the point, as described below. The pilot reference signal (or pilot pilot frequency) is a pilot that transmits all antennas on the eigenmodes of the ΜΙΜΟ channel. The user terminal can transmit as follows - the upper γ & the custom key is guided by the reference signal: ^up,m Υ — V* Λ ” Equation (6) 99590.doc •14· 1361601 where n is one corresponding to a pilot symbol transmitted on an eigenmode w of the guided reference signal;

2Lup,m is a transmission vector corresponding to the uplink guided reference signal of the eigenmode W; and the wth eigenvector or row of the system, where 厶 = [heart u. The received uplink guided reference signal at the access point can be expressed as: = ΗϋρΪΒρπ + 5叩=HupEut:^;^ +2„p

+Snp=test. a h +flup ' Equation (7) where HuP,m is a reception vector corresponding to the uplink guided reference signal of the eigenmode w; ~the first of the S; n diagonal elements; and u Λ

(4) The wth eigenvector or row of the system, where Bean·ρ =υ如...心~]. Equation (7) shows that 'in the absence of noise, the received uplink guided reference signal at the access point is approximately equal to the access point. Various estimation techniques can be used, based on the guided reference of the manufacturer's terminal material. The signal is used to obtain an estimate of the uplink channel response. In the embodiment, it is obtained. The estimated value, firstly, the received vector U is multiplied by the pilot (4) and then integrated for each eigen, a plurality of received guided reference symbols to obtain a one-way-:, the vector corresponding to Estimated value of eigenmode. Since the eigenvector of the temple has unit work to force sheep, the eigenvalue of the eigenmode 99590.doc -15- 1361601 can be estimated according to the received power of the pilot reference signal of each eigenmode. Each of the ~ elements of the estimate of 'cr'U' is obtained by dividing one of the elements of ^. In another embodiment, the _MMSE technique is used to obtain an estimate based on the received vector ‘. Since the pilot symbol ~ is known, the access can be used to derive the estimated value so that the received pilot symbol nine (which is obtained after performing matched filtering using the received vector km) and the seven-sense The mean square error between the frequency symbols ^ is minimized. The access point can be performed on the estimated value of (where m = 1 .. Ns). ★ Example * 5 ′ φ is used to obtain an estimate for each eigenvector, such as ~. The eigenvector estimates may not be orthogonal to each other due to, for example, noise in the received reference signal, changes in the MIM channel response, and the like. Therefore, the access point can perform on the one eigenvector estimate

Gram-Schmidt is orthogonalized to obtain orthogonal transmit vectors. In summary, the access point obtains an estimate of the transmit matrix t, and & is derived from the user terminal based on U. The access point then uses this transmit matrix to perform spatial processing on the downlink transmission. 1. Post-performance calibration The corrections obtained from the initial calibration and/or may include errors caused by various sources such as: (1) Channel estimates for initial calibration

Hdn and Hup are not ideal, (2) access ^fj; 3 1) The transmit/receive chain at the point of the user and the user terminal is changing' and so on. The error in the correction matrix will be found in the following two cases: (1) the transmission matrix used by the user terminal for spatial processing and self-exporting... and (7) by the access point for spatial processing and self-contained 99 99990.doc • The uplink of the 16·1361601 is transmitted by the pilot reference signal. Improved performance can be obtained if the error in the correction matrix can be estimated and eliminated. - The access point and/or the user terminal can perform subsequent calibration to estimate the error in the correction moments, the arrays I and I. For the sake of clarity, the following is a description of the subsequent calibration performed by the user. In subsequent calibrations performed by the user terminal, the access point uses a correction matrix to transmit a pilot on the downlink, and also uses the transmit matrix & and the correction matrix inverse p to transmit on the downlink φ path. Guide the reference signal. The downlink guided reference signal can be expressed as: where. Similar to the above, as described for the uplink piloted reference signal, the user terminal can obtain an estimate based on the received downlink referenced signal. For the sake of brevity, the estimated value of L纩 derived from the downlink guided reference signal is referred to as an “actually received” transmit matrix, which is an informal one containing the estimated value of Xut and the estimated value of the left. Matrix. (For clarity, the "~" above the matrix indicates that it is an unnormalized matrix". The user terminal also obtains another version of the packet based on the pilot transmitted by the access point. ~ η Correction moment The error of 3^ap and the box can be represented by the calibration error diagonal matrix ^ and L respectively. Therefore, the correction matrices I and I can be expressed as: heart = 匕1 and £W=K. Equation (8) The error in the correction matrix is small, close to the complex value of 1 + y 。. Therefore, the value ή α ΐ η can be expressed as · · The exemption of the diagonal element of the line m is calibrated and the downlink channel response is estimated 99590.doc 17 丄 601 Equation (9) or

7Matrix K and L contain the "true" errors in the heart and I, respectively. Can the % or the guess value or the estimated value be marked as 込? And 仏^. A hypothetical J downlink channel can be defined as: , = bin <. equation (10)

The hypothetical downlink channel is "one of the conjecture values, which is derived under the assumption that the applied t corrects the error matrix of the correction matrix U. If the ideal guess value of the center of the equation (10) is The ideal estimate of ^ in the formula (9), then the heart and c = heart. -n The spatial processing at the access point can be expressed as: Equation (1))

The Eighth L system is obtained from the singular value decomposition of the packet dn, and the g coffee is obtained from the downlink ΜΙΜΟ pilot. The user terminal does not have a value of I, but only has its guess value. Therefore, the user terminal calculates the denormalized emission matrix L'. If the calibration error matrix is 'and 仏, the irregularized emission matrix is obtained from the access point as shown in the following equation by the imaginary method. (10) The ideal guess value of the right iip system and the general ideal guess value of the system, then the equation (j2) is equal to the equation (1). If this is the case, then red P. The "post" user terminal performs the processing in the same manner as the access point should be processed in response to the received uplink key being guided by the reference k number, and 99590.doc 8 -18- !3616〇1

The resulting "emission matrix Hg" is obtained, which is a normalized emission matrix similar to packet p. For example, an access point may perform a Gram Schmdit positive affiliation on the received eigenvector to improve the performance of its transmit steering vector. In this case, the user terminal will perform the same normalization on the eigenvectors in & The user terminal only simulates the processing normally performed by the access point and the user terminal, although it is assumed that the calibration error is represented by the heart and the 矩, and the 仏 仏 should be used by the access point to transmit the downlink key. The reference signal and spatial processing are performed on the downlink transmission. The spatial processing at the user terminal can be expressed as: Equation (13) Similarly, the 'user terminal does not have a store' and only has its guess value red. The user terminal is calculated for this by itself - assuming the emission matrix

Yhyp - Hhyp § apiig. Equation (14) Equation (14) is equal to equation (13) if the ideal guess value of 迅 &< and 仏P#s:>p is the ideal guess value. The matrix is an unnormalized matrix comprising - a user terminal transmit matrix corresponding to the access point transmit matrix 仏) and a diagonal matrix with a heart-like assumption that the user terminal has received the matrix by t(1) The user terminal uses the exemption, transmits a subscription link to receive the reference signal, and (2) the access point performs the normal processing on the received uplink signal by the pilot reference signal to derive its transmission matrix, and saves The point is 仏 transmit-downlink is subject to (iv) reference signal, (4) repair, the errors of matrix Kap and κω are represented by 仏 and 仏 matrices, respectively, and (7) assume that there is no channel estimation error for pilots from downlink milk. 99590.doc -19- 1361601 Equations (12) and (14) are correct if the calibration error matrices 込p and ubiquitous u t correctly represent the true error of the modified matrix mouse and L respectively. From the downlink link:: The actual received transmit matrix obtained by the $Guide reference signal and the difference from the downlink link: the 发射 pilot obtained by the assumed transmit matrix can be calculated as follows: ' Ε=ν.-ν ^, one program (15) where the system - the error between the brother and the child constitutes a Nap order matrix. What is the error matrix 1? And 込, the accuracy of the guess value. A variety of adaptive procedures can be used to adjust the matrix 込P and to drive the error matrix E towards zero. These adaptive procedures include the MMSE Adaptive Procedures and the steepest declines. For adaptive programs, the diagonal elements of the heart and the 仏 can be initialized to 1 + y_〇. For the MMSE adaptive program <^·τ, the approximate partial derivative of the element in 仏 relative to the elements in 叩 and t t is calculated. If the "first" τ (the uppermost element) of & is set to 丨.0 by initial calibration, there is no need to adjust the element. In addition, the error matrix 1 is not affected by the values of the elements in the frame. Thus, for example, t can be normalized by defining the real component of the first element of 仏t as 1.0. Further, an eigenvector (or any phase) can be converted by a complex number of unit values, which does not affect its effectiveness. Therefore, a set of phases can be selected to avoid proximity without losing generality. • This property allows the conversion of 仏 using an arbitrary unit value factor... thus • The imaginary component of the first element of 仏 is defined as 〇.〇. The MMSE adaptive procedure can be performed as follows. Let Jiang be a real and imaginary component of each diagonal element of length 99 (Nap+Nut-2) and 99590.doc •20· 8 other than the first element of ubiquitous and Q_ut except I Real vector. Vector can be defined as:

2 ... N ap r 92(iD =Im{!2ep(z.,i)}, where ζ· = 2 Ν 2 («η = Re(2Ui(z,2)}, where ζ· =: 2 ν , 一议}, where, = 2...Ν::, where the Η 元素 element of ❼ is a; 2βΡ (Ζ·, Ζ·) is the brother of Hap ί a diagonal element; and δ« /(Ζ·,/) is the ίth diagonal element of 仏t. The element of life with an odd number corresponds to the real component of the diagonal element in red and 仏, and the element with the even number attached Then corresponds to the imaginary component of the diagonal element in 仏Ρ and 仏. Before the birth, 2Nap_2 elements correspond to the Nap-1 diagonal elements in the red except for the first _ prime ,, and the last of the birth The "two elements of the muscle correspond to the NurHgJ diagonal elements of the Q_ut except for the first one. π ^ is a real vector of length 2NapNut and composed of real and imaginary components of the elements in i. Defined as: =Μ^} ' where /=1 υ) where the first element of the system; and • ugly (by) the first in the system..., _; • the element at the line < . The elements with odd-numbered labels correspond to the elements of the J-marked elements in the Ε, which correspond to the imaginary components of each element in £ The error matrix 爻0 1...N, N • · _ Kao and the real component can be obtained by using the inward a Khan equations (10), (12), (1 sentence and (15), and the anti-medium band 99590 .doc 1361601 For MMSE adaptive procedures, the division of τo in 込p or 込t can be disturbed: and the calculation is defined by equations (1〇), (12), (14) and 〇5) • The function 'to produce the partial or imaginary component of the element in & relative to the partial derivative of the real or imaginary component of the prime in (1). As part of the calculation of k-, A single # term can be chosen to minimize 氐. This will normalize the phase of the first element of Q_ut. The approximate partial derivative of the element in e relative to the element in red can be expressed as: Φ ^ 6 '卜1 ... 2(Nap+Nut-2) and _/=1 ... 2NapNut Equation (16) where Aj is a vector of length 2 (Nap+Nut-2) containing δ at the yth element The small real value, including zero at other positions; and the approximate partial derivative of the /e element in the system relative to the third element in the Liaozhong. The approximate partial derivative can be obtained as follows. According to the mountain=江+4, the stem flj is calculated. Then, the function defined by equations (10), (12) and (14) is evaluated to obtain the function (which includes 込Ρ, and 仏t, ,). A new (or "repaired") assumed emission matrix is then subtracted 5 〆 from 乞 to obtain a new error matrix m, which is used to form a new erroneous i vector w _ , from the first element, the element (marked as woven a+4i in equation (16)), subtracts the yth element of L (marked as Ei in equation 1 (16)). Then, divide the result of the subtraction by the value of A', and perform equations (10), (12), (14) for each of the 2 (Nap+Nut-2) elements of SL. (15) Calculate to obtain a corresponding new error to 99590.doc -22· 1361601 §j·. For each new error vector l·, the 2NapNut elements in argon are subtracted one by one from 2NapNut elements in t to obtain 2NapNut approximations of 2NapNut elements in L relative to the first element in the birth. Partial derivative. A matrix of 2NapNut X 2 (Nap+Nut-2) may be formed by all partial derivatives corresponding to all elements of Lan and Yu. Each of the lines contains 2NapNut approximate partial derivatives of 2NapNut elements in e relative to one element in Liao. The 2 (Nap+Nut-2) lines in A correspond to 2 (Nap+Nut-2) elements in a. If the relationship between equations (10), (12), (14), and (15) is approximately linear, the estimated difference between the calibration error guess value and the actual calibration error can be expressed as: " Equation (17) Where ri is an update vector corresponding to the difference between the estimated Liao and the actual calibration error. The update vector fork has the same format and dimension as the vector, and the real vector consists of the real and imaginary components of the diagonal elements other than the first element in the 叮t and 仏t. If a is not a matrix (which is usually the case), then there is no simple inverse matrix. At this point, the opponent program (17) uses the 4th Moh-Penrose (M_e_ Penr〇se) virtual inverse matrix. The virtual inverse matrix is only a matrix satisfying the equations AA A-A and a 4A_1=A_1. The virtual inverse matrix can be generated by performing singular value decomposition on the following method, and then calculating the virtual inverse matrix by Δ ,, wherein the _ is a corresponding non-zero diagonal in the pan The inverse of the elements constitutes the diagonal matrix. 99590.doc

-23· 1361601 The calculation of the partial derivative matrix 4 assumes that the functions defined by the equations (ίο) to (丨3) are approximately linear for the calibration error of the evaluated magnitude. Because of this

The linear assumption is not completely accurate, so the program can be iterated multiple times to determine the correct calibration error. In some cases, the program does not converge. However, convergence can usually be achieved by simply selecting a different initial error value for the calibration error. If convergence is not achieved, the user terminal may also obtain another version of ^ and In according to another estimated value of the downlink pilot reference signal and the downlink pilot, and perform the use of the new matrix. MMSE adaptive program. If equations (10), (12), (14), and (15) are linear, then the mean square of each element in the blue is minimized. However, since the equations are not linear, the lamp is used instead of red and the procedure is repeated. Thus, the calibration error vector can be updated as follows: Equation (18) where "the number of iterations is attached; the calibration error vector used for the „th iterations; the update for the obtained iterations Vector; and the calibration error vector used for the (n+1)th iteration. The above calculation can be repeated for any number of iterations. Each iteration uses the updated calibration error vector obtained from the previous iteration. When the vector is updated (the illusion is small enough, the program can be ended. For example, the end condition can be I丨Ζ(»)丨丨, where the sum of the squares of the values of the elements in ii(n) is a threshold. As another example, the end condition is 99590.doc

-24- 1361601 can be brother _<~(for all!·), where the first sleeve of the force system 1(„) is 70, and the other value is. After all the iterations have been completed, The final estimate matrix of calibration errors is expressed as QaWne/ and Q_ut, y7„e/. For the steepest descent adaptive program, a total error can be defined as: ζ=_2=ιι乞-Lu2. Equation (19) The total error z is obtained by summing the squares of the values of the elements in E. 2 The partial derivative of each element in a can be calculated as follows: 5 2 2 (g + l) - z (q) dg > δ , where 卜 1...2 (NaP + Nut '2), equation (2 〇) where ^ is the approximate partial derivative of z in relation to the third element in a, which is a vector of length 2 (Nap+Nut-2), which contains a small real value of j at the third solid element It contains zeros at other locations. The approximate partial derivative heart can be obtained as follows. First press • to calculate a vector mountain. Then, the functions defined by equations (10), (12), (Μ), and (15) are evaluated for the mountain to obtain a new error vector, and then, for the new error vector. To calculate the total error h, as shown in equation (19). Then, subtract the total error z obtained using red from the total error A obtained by using the tie (marked as 2(a+4) in the square (20)) (in equation (2〇) it is marked as z(w) ^ Then divide the result of the subtraction by the gain to obtain the heart. Repeat the calculation for the 2 (Nap+Nut_2) elements in the middle. Approximate for the 2 (Nap+Nur2) elements for Liaozhong The partial derivative forms a 2(Nap+Nut-2) dimensional vector £. Each element in the & is the slope of the total error evaluated at the corresponding element in Liao. Then, the calibration error vector can be updated as follows: Equation (21) 99590-doc -25- !3616〇1 where &(n) is the slope vector of the „th-stacked iteration, U(«) and

IrfO + l) is the calibration error vector of the second and second iterations in the steepest descending process. Any number of iterations can be repeated for the above-mentioned 呻 殳 殳 leaf. Each iteration uses the updated calibration error vector a^(«+i) obtained from the previous iteration. The program is terminated when the total error 〇钧 2 疋 is small enough, for example, less than a threshold 訏'. Two adaptation procedures have been described above for the estimation of the actual calibration error. Other adaptive and non-adaptive procedures may also be used, which are still within the scope of the present invention. The user terminal can update its correction matrix to compensate for the calibration error as follows: Κδ:- 〇 Equation (22) As shown in Figure 2, the user terminal uses the new correction matrix i instead of the previous revision matrix to transmit the uplink. Perform space processing. The user terminal can be used. The Wu difference matrix Qap, /infl/ is sent to the access point, and then the access point can update its correction matrix to ^aiyiw=:Z'. To reduce the amount of transmission, if the calibration error matrix P, /ina/ satisfies a predetermined threshold, the user terminal can only send the reference error matrix up to .... Figure 3 shows a process performed by the access point and the user terminal for initial calibration, positive: operation, and subsequent calibration. First, the access point and the user terminal perform an initial calibration to calibrate their transmit and receive chains and derive the L positive matrix print and 1 (block 3 10). The initial calibration of Yanhai will be explained below. After that, in normal operation, the access point uses its correction matrix ap to transmit - 99590.doc -26 - 1361601 downlink ΜΙΜΟ pilot (block 322) ^ the user terminal obtains according to the downlink ΜΙΜΟ pilot A calibrated downlink channel response estimate of 0cdn (block 3 24) is performed and singular value decomposition is performed on &dn to obtain its emission moment: argon (block 326). Then, the user terminal uses t and transmits an uplink-guided reference signal as shown in equation (6) (block 328). As described above, the access point receives the uplink guided reference signal and derives its transmit matrix L (block 330). The access point and the user terminal use the transmit matrices t and t for spatial processing, respectively. In subsequent calibrations, the access point transmits the downlink guided reference signal using ip and upper p, and further transmits the downlink MIM0 pilot using t (block 342) as described above (block 344). The user terminal derives an actual denormalized emission matrix £ according to the downlink guided reference signal. The user terminal also calculates the denormalized emission matrix t according to the equation (1〇) and (12) or « according to the modulating downlink channel response estimation value ficdn and the calibration error matrix of the transmission matrix. φ 346). The user terminal is further processed (e. g., performing orthogonalization) in the same manner as the access point should perform to obtain a normalized transmit matrix instant (block 348). Then, the user terminal calculates the assumed denormalized emission matrix sibling block 35(1) according to the emission matrix 仏 and the calibrated downlink channel response estimate gedn according to equation (1〇) and its system. If the access point uses 仏 to transmit the downlink guided reference signal, then the matrix 系 is the unnormalized transmit matrix that the user terminal should have received. The user terminal then modifies the calibration error matrices 込p and ilat according to the transmit matrix 交 and the coffee (block 352). Blocks 346 through 99590.doc -27- 1361601 352 can be performed using an adaptive program. Thereafter, the user terminal can update its correction matrix capacitance using the calibration error matrix (block 354), and thereafter the access point can update its correction matrix ap using the calibration error matrix (block 356). Figure 4 shows an adaptation 400 that can be used for blocks 346 through 352 shown in Figure 3. First, the assumed transmission matrix Y (4) is calculated from In, the general ap, and the vertical ut (block 410). Block 41A corresponds to blocks 346 to 35A in FIG. Next, as shown in equation (15), the error matrix is calculated [as the difference between the transmission matrix L and 〜 (block 412). Then, as shown in equation (16), each element in the derived error matrix K is associated with the calibration error matrix to know each element in the selected element of the mind (eg, all diagonal elements except the first element). The partial derivative (block 414). As described below, for ease of calculation, the matrix and the matrix and the ut can be placed in a vector form. As also mentioned above, partial derivatives can be derived for the real and imaginary components of the elements in the matrices, respectively. Next, as shown in equation (17), the update vector r is calculated from the partial derivative matrix 丄 and the error matrix (block 416). ‘Then, as shown in Equation ^, the updated vector sense is used to update the calibration error matrix and W block 418). Next, the update vector is found to satisfy the end-end condition (block 420). If the answer is "yes", the process_ ends. Otherwise, the process returns to block 410 and performs another iteration. Figure 5 shows a steepest descent adaptive procedure 5 〇〇 which can also be used for blocks 346 to 352 of Figure 。. First, Tong Chenghe a ~ calculates a hypothetical emission matrix Yhyp based on HU and pan ut (block 510). Next, go to +, 1~~ Next, as shown in equation (19), take the total error 2 as a calculation (block 5i2)n, as shown in equation (2〇), and derive the total error relative to the calibration. Error matrix red and 99 99590.doc •28· 1361601 The partial derivative of each of the selected elements (block 514) 〇 can be used to form the matrix ilap and 仏 as vectors, and for these θ „ , ν θ 丨 〒 实 之 之 之 之 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚 虚The total error 4 is satisfied - the end condition (block 518). If the answer is yes, the process ends. Otherwise, the process returns to block 5 10 and another iteration is performed.

In the above description, the user terminal estimates the calibration errors in both the correction matrix and the ^. To simplify subsequent calibration, the user terminal can assume that the correction matrix does not contain errors, but only the errors in the correction matrix t. This is equivalent to setting the calibration error matrix 込p to an identity matrix. By omitting 込〆, the vectors £, 1, and & and the dimension of the matrix 4 can be reduced, so that this can be greatly reduced by P. For the sake of clarity, the above description is directed to the case where subsequent calibration is performed by the user terminal. The access point can also perform subsequent calibrations. In this case, in Figure 3, the access point and the user terminal switch roles. At this point, the user terminal will transmit an uplink pilot reference signal and an uplink MIM 〇 pilot 'and the access point will perform the calculation to derive the public and ut » also for clarity, the above The subsequent calibration is illustrated for a single carrier PCT system. Subsequent calibration can also be performed on a multi-carrier system that can utilize orthogonal frequency division multiplexing (〇FDM) or some other multi-carrier modulation technique. OFDM effectively divides the overall system bandwidth into multiple (Nf) orthogonal sub-bands, also commonly referred to as tones, sub-carriers, frequency bands, and channels. For OFDM, each subband is associated with a corresponding carrier that can be modulated using 99590.doc • 29-8 1361601. For a 〇 〇 system (MIM0_0FDM system), the above calculation can be performed for each of a plurality of sub-bands. Since there may be a high degree of correlation between the frequency bands in each of the neighboring states, it is possible to perform the calibration 'by example' in such a manner as to reduce the number of sub-bands for performing subsequent calibration, accelerate convergence, etc. 2 Initial calibration

To implement the initial calibration to derive the correction matrix and I, an effective uplink channel response estimate value n and the effective downlink channel response estimate ^up are obtained by an entity (/user terminal or access point). The channel estimate values ^ and 0up can be obtained from the downlink and uplink ΜΙΜ〇 pilots, respectively.修正 The correction matrix can be calculated based on the 0up, ❹ matrix ratio calculation or μ job. In the calculation of the matrix ratio, first calculate a 阶 χ 阶 阶 £ £ : : : : : : : : £ £ £ £ £ , , , , , , , , , , , , , , , , , , , , , , , , , Each diagonal element in the correction matrix goldprint of the access is set equal to the average of the normalized columns in ^. First, by multiplying each of the Nap elements in each column of £ by the first element in the column, the column: normalizes...w爰 normalizes the column by 5 hai (by a vector) The average value of the mutual coffee is calculated as the sum of the normalized columns divided by Nut. Then, set the ~ diagonal element in your mind to equal! , the Nap elements in the middle. Since 99590.doc -30. ουι is normalized, the first element of I is equal to the correction of the user terminal ^ " The diagonal element of the car **_ is set equal to £: the inverse of the rule The average value. First, the row is normalized by multiplying each element in (10) 0. a...aP) by the 7th diagonal element in I. Then, the inverse of each normalized line is calculated as follows: Mean value (represented by -vector ^ is wide (1) to find the inverse of each normalized line, where the system is inversed by S-S, and (7) is obtained. The inverse normalization line ^, and (3) divide each element in the result row by , to obtain. The N diagonal elements of I are set equal to ΚΙ 元 ^ Ut in the MMSE calculation towel, is valid The downlink and uplink channel response estimates iih^iLp derive the correction matrix box and 1 to minimize the mean square error (deletion) between the calibrated downlink and the uplink link response. This condition can be expressed. It is: Equation (24) min |(fiU Eip )r - H 丨, which can also be written as: min where sap is a pair of angular matrices, thus fluorine. Equation (24) is set to equal the first element of the following constraints If there is no such constraint, it will be obtained that all the elements in the matrix & and 氪 are set equal to zero. ^ τ The solution. In equation (24), first obtain a matrix Y: X=3^PSL- H. Connect - Γ also ... θ - Next 'get every sacred NapNut element The square of the value. ± Shandingwan This 'mean square error (or square error, which is not divided by NapNut) is equal to the sum of all squared values. 99590.doc • 31 - AUUi mmSE calculation is performed as follows. For the sake of brevity, mark each element in i as (4), mark each element in aup as (4), and mark the diagonal element in the branch as {the diagonal element of i will be marked as {vy} ' where / = 1 N and / = 1

Aap and 7 i.-.Nut. The mean square error can be rewritten as follows according to equation (24): H 丨, w, equation (26) is the same as ', = 1 constraint. The minimum mean square error can be obtained by taking the partial derivative of equation (25) with respect to ... and setting the partial derivatives to zero. The result of the operation is as follows: Σ(αί/ΊΉ=〇ν , where i-2...Nap, and equation (26a), where /-l."Nut. Equation (26b) is in equation (26a) In the case, Ml = 1, and thus there is no partial derivative in this case, and the attached standard is from 2 to Nan. The set of Nap+Nut-1 equations in the equations (26a) and (26b) of the equation can be It is more conveniently expressed as the following matrix form: Equation (27)

Bk=z, where 99590.doc -32· 1361601 'ΪΜ2 household 1 0 ... 0 ^^2\a2i ... 0 ϊ\%\2 ο y=j ··· ... ... 0 ... 0 N , 5 = 0 ... 0 尸1 -αιΑι ... Ά, ΪΜ 0 ... 0 ... ... 1=1 0 ΪΜ 〇isl ... _β2Ν„Λ·Νβ ... 0 0 ... ... 0 « « 2 ' 0 ' «3 0 and?= 0 V1 α\Φη V2 β12612 »1Nm The matrix contains Nap+Nut-1 columns' where the front Nap_H@ column corresponds to the Nap-1 equation from equation group (26a), The last Nut column corresponds to the Nut equation from the equation group (26b). The elements in the matrix and in the vector 2 can be obtained according to the elements of ^up. The diagonal elements of p and box (2) Included in the vector, the vector K can be obtained as follows: 匕 = Ί. Equation (28) The result of the MMSE calculation is the correction matrix (and ^, which will result in the calibrated downlink and uplink channel responses. The mean square error is minimized, as shown in equation (24). 3. Spatial Processing Table 2 summarizes the eigenmodes performed by the user terminal and the access point for the pass 99290.doc -33 - 1361601 Transmitting and receiving data Space processing. Table 2 Uplink Key Downlink User Terminal Access Point Transmit: Receive: Receive: ηηι Λ Yes · Transmit: Xdn - SapHap&n For the embodiment shown in Figure 2 and Table 2, Correction Matrix & And & are applied separately

On the transmitting side of the access point and the user terminal. 4. System Figure 6 shows a block diagram of one embodiment of an access point 11 and a user terminal 150 in a TDD system. On the downlink, at the access point u, a transmit (τχ) data processor 610 receives traffic data (ie, information bits) from a data source 6 〇 8 and receives signals and other data from a controller 63 〇 . The data processor 61 formats, encodes, interleaves, and modulates (or symbolically maps) different types of data and then provides data symbols. As described in this document • “data symbols” correspond to the modulation symbols of the data, and “pilot symbols” correspond to the modulation symbols of the pilots. The pilot symbols are known a priori for the access point and the user terminal. A spatial processor 620 receives data symbols from the data processor 610 'performs spatial processing of the data symbols, multiplexes them into pilot symbols as needed (eg, for channel estimation, calibration, etc.) and then provides The transmitted symbols flow to Nap modulators (MOD) 622a through 622ap. Each modulator 622 receives and processes a corresponding transmit symbol 々, L ' to obtain a corresponding OFDM symbol stream, and the corresponding OFDM symbol /;, L is further processed by a transmit chain within the modulator to Obtain a corresponding 99590.doc •34- 1361601 line link modulated signal. Then, from the Nap antennas 624a to 624ap respectively, the downlink modulated signals from the modulators 622a through 622ap are transmitted. At the user terminal 150, the Nut antennas 652a to 652ut receive the transmitted. The downlink link modulated signal, and each antenna provides a received signal to a corresponding demodulator (DEM〇D) 654. . Each demodulator 654 (which includes a receive chain) performs processing complementary to that performed at modulator 622, and then provides received symbols. Then, a receive (RX) spatial processor 660 performs receiver spatial processing on the received symbols from all of the Nujsi demodulator 654 to obtain detected symbols. The detected symbols are transmitted by the access point. Estimation of the data symbol „·(·value β for subsequent calibration, RX spatial processor 66〇 provides (1) a calibrated downlink channel response estimate dn, which is transmitted from an access point The link ΜΙΜΟ pilot is obtained, and (2) corresponds to a received symbol of the downlink guided reference signal transmitted by the access point. An RX data processing 670 process (eg, symbol de-mapping, de-interlacing, and decoding) The detected n number 'and provides decoded data. The decoded data may include • recovered data, signals, etc., provided to a data slot 672 for storage and/or to a controller 680 for further processing. The uplink processing may be the same as or different from the downlink processing. The data and signals are further processed by the spatial processor 692 after being processed (e.g., encoded, interleaved, and modulated) by the data processor 69. At The multiplexer is multiplexed with the pilot symbols to obtain the transmitted symbols. Then, the transmitted symbols are further processed by the modulators 654a through 654Ut to obtain an uplink modulated signal 'then, then these Nuts are uplinked The link modulated signal is transmitted to the access point via Nut antennas 652a through 652ut. For the implementation of the above-described implementation 99590.doc - 35 - 1361601, the 'user terminal 15 发送 sends back the correction matrix t for initial calibration' and The calibration error matrix 仏ρ//ηβί can be sent back for subsequent calibration. At the access point 110, the 'uplink modulated signal is received by the antenna 624, demodulated by the demodulator 622, and then The rx spatial processor 640 and an rx data processor 642 perform processing in a manner complementary to the processing performed at the user terminal. The RX data processor 642 then provides the matrix to the controller 630. φ is to perform an initial calibration, The access point and the user terminal transmit MIM 〇 pilots on the downlink and uplink respectively. Each entity derives an effective channel response estimate for its link. One of the entities (eg, an access point) will pass The channel estimate is sent to another entity (such as a user terminal) for calculating the correction matrix boxes of the two entities "and". The entity that derives the correction matrix uses its correction matrix and sends another correction matrix Back to another entity. To perform subsequent calibration, one of the entities (such as an access point) transmits both the piloted reference L number and the ΜΙΜΟ pilot. The other entity is derived based on the received pilot, • as described above The calibration error matrix and iW", the entity that derives the calibration error matrix will use its calibration error matrix' and can send another calibration error matrix back to another entity (eg, if the error is large enough, control states 630 and 680 control the access point, respectively) And the operation of various units at the user terminal. The controller 63 〇 and/or may also perform initial and subsequent calibration processing (eg, calculation of the correction matrix 匕 and t and the calibration error matrix 込 and 込. The memory units 032 and 682 store the data used by the controllers 630 and 680, respectively). And the code. The calibration techniques described in this article can be constructed by various methods. In terms of terms, 99590.doc 8 -36· These technologies can be built on hardware, construction schemes, used to perform initial and:: two... Built in or multiple application-specific integrated circuits (ASIC), digital signal (4)

Bit 唬 Processing Unit (DSPD), Programmable Logic Device (R), Field Programmable Closed Array (FPGA), Processor, Control I Micro Control n, Micro Processing H, and other materials used to perform the functions described herein The electronic unit, or a combination thereof.

For software construction scenarios, modules (e.g., programs, functions, etc.) capable of performing the functions described herein can be used to construct such calibration techniques. The software code can be stored in the _s memory unit (for example, the memory unit 632 and the milk in FIG. 6): and the processor can be executed by the processor (for example, the controller (10) and the caller 1 memory early reduction can be built on the processor The internals may also be external to the processor, 1 being built external to the processor, which may be communicatively coupled to the processor via various methods as are known in the art.

The headings included in this article are intended to facilitate access and help determine the location of certain chapters. These headings are not intended to limit the scope of the concepts described under the headings, and the concepts may apply to other sections throughout the specification. The above description of the disclosed embodiments is intended to enable any person skilled in the art to make or utilize the invention. The various modifications of the embodiments are readily apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but rather to the broadest scope of the principles disclosed herein. [Simplified description] 99590.doc -37· 1361601 Figure 1 shows the transmission and reception parts of the access point and user terminal in the TDD system; Figure 2 shows the correction matrix using the access point and the user terminal. Compensating for its transmit/receive chain; Figure 3 shows a process performed by the access point and user terminal for initial calibration, normal operation, and subsequent calibration; Figure 4 shows a minimum mean square error (MMSE) adaptive procedure;

Figure 5 shows a steepest descent adaptive procedure; and Figure 6 shows a block diagram of the access point and the user terminal. [Main component symbol description] 100 TDD ΜΙΜΟ System 110 Access point 112 Correction unit 114 Transmit chain 116 Antenna 124 Receive chain 150 User terminal 152 Antenna 154 Receive chain 162 Correction unit 164 Transmit chain 608 Source 610 Transmit (ΤΧ) Data processing 620 ΤΧ Space Processor 99590.doc -38* 1361601 622a Modulator 622ap Modulator 624a Antenna 624ap Antenna 630 Controller 632 Memory 640 RX Space Processor 642 RX Data Processor

644 data slot 652a antenna 652ut antenna 654a modulator 654ut modulator 660 RX space processor 670 RX data processor 672 data slot

680 Controller 682 Memory 688 Source 690 TX Data Processor 692 TX Space Processor 99590.doc -39-

Claims (1)

1361601 X. Patent Application Range: 1. A method for calibrating a downlink and an uplink channel in a wireless multiple input multiple output (MIM) communication system, comprising: according to a transmitting entity and a receiving entity a first pilot received by one of the channels, deriving a first transmit matrix; deriving a second transmit matrix based on a channel response estimate and the first and second calibration error matrices, the mean channel response estimate The _ estimated value of the response channel, and the data is derived from the second pilot received via the ΜΙΜΟ channel, the first calibration error matrix including a response for compensating the transmit and receive chains at the transmitting entity An estimate of the error in the first correction matrix, and the second calibration error matrix includes an estimate of an error in a second correction matrix for compensating for a response of the transmit and receive chains at the receiving entity; The second emission matrix modifies the first and second calibration error matrices. 2. The method of claim 1, wherein the first pilot is a guided pilot frequency received via a plurality of eigenmodes of the chirp channel. 3. The method of claim </ RTI> wherein the second pilot is a ΜΙΜΟ pilot formed by a plurality of pilot transmissions transmitted from a plurality of transmit antennas at the transmitting entity, wherein each of the transmit antennas The pilot transmissions are all identifiable by the receiving entity. 4. The method of claim 1, wherein the deriving a second emission matrix comprises singulating the ΜΙΜΟ channel response estimate to obtain a first eigenvector matrix of the μιμΟ channel, 99590.doc based on the ΜΙΜΟ channel response estimate And the first and second calibration error matrices, calculating a second eigenvector matrix of the one of the chirp channels, and selecting the second emissive matrix from the second eigenvector matrix and the μιμ〇 channel response estimate. 5. The method of claim 4, wherein the deriving a second transmit matrix further comprises processing the second eigenvector matrix to obtain a third eigenvector matrix 'where the processing of the second matrix is performed by the transmitting Entity executed by the entity to generate a transmit matrix according to the received pilot frequency received by the transmitting entity from the receiving entity, and wherein the third eigenvalue is calculated according to the I matrix and the channel response estimate The second emission matrix. 6. The method of claim 5, wherein the processing the second eigenvector matrix comprises performing orthogonalization on the eigenvectors in the second matrix to derive orthogonal eigenvectors of the third matrix . The method of claim 1 is wherein the first and second calibration error matrices are modified according to a minimum mean square error (MMSE) suitability procedure. 8. If requested, the first and second calibration error matrices are modified, and the leaf computing p error matrix is used as the difference between the first and second emission matrices, to: The partial derivative of the element in the 5 Hz error matrix relative to the selected element in the first and second calibration error matrices, 2 calculated according to the partial derivative and the error matrix - updating the vector, and updating with the update vector The first and second calibration error matrices. 99590.doc 9. The method of claim 8, wherein the deriving the partial derivative comprises deriving a modified second emission based on the chirp channel response estimate, the first and second calibration error matrices, and an error vector The matrix, the modified error matrix is used as the difference between the first emission matrix and the modified second emission matrix, and the partial derivatives are calculated according to the "comparison matrix" and the modified error matrix. For example, in the method of claim 8, the sixth-mile error matrix and the first and second schools: the matrix includes complex-valued elements, and each of the complex-valued elements has a - - and a virtual branch, and wherein The partial derivatives are derived for the real components and the quantities respectively. 11 · If the method of the method 8 is used, the basin is in the step of including, and the middle (four) is changed to use the first and second calibration errors. The partial derivative forms one and the partial derivative matrix is inversely calculated: (d) and wherein f is according to the error matrix 12. According to the method of claim 8, wherein: the new element is the element of the tenth and the first In the second calibration error matrix, except for the top left Side element: includes the first and second calibration error moments 13. As in the method of claim 8, all diagonal elements of the political matrix. The matrix advance step includes the modification of the first and second calibration errors. "Hui juice - error matrix, export ^ quantity, and update the first - and ... derivative, calculate an update to the update vector to meet - end ^ difference material multiple times, until the 99590.doc 1361601 14, as in the method of claim 1 In the second place, the first and second calibration error matrices are modified according to a steepest descent adaptive procedure. The method of claim 1, wherein the modifying the first and second calibration error matrices comprises: a ten error matrix as a difference between the first and the second transmit matrix, and calculating a total error according to the error matrix, and deriving 5 The partial derivative of the total error relative to the selected one of the first and second calibration error matrices, and the first partial second calibration error matrix is updated using the partial derivatives. 16. The method of claim 15, wherein the total error is calculated as the sum of the squares of the element values in the error matrix. 17. The method of claim 15, wherein the modifying the first and second calibration error matrices further comprises: repeating the calculating an error matrix, calculating a total error, deriving a partial derivative, and updating the first and second calibrations. The error matrix is repeated multiple times until the total error satisfies an unconditional condition. 18. The method of claim 1, further comprising: updating the second correction matrix using a second calibration error matrix. 19. The method of % 求, wherein the first correction matrix is updated using the first calibration error matrix. 20. The method of claim 1, wherein the receiving system is a user terminal in a time division duplex (TDD) MIMO system, and the transmitting system is an access point. 21. The method of claim 1 wherein the system utilizes orthogonal frequency division multiplexing 99590.doc 1361601 (OFDM)' and wherein the first and second pilots are received based on the plurality of bits, and The second calibration error matrix. Each of the sub-bands derives a set of devices in the sub-band for the sub-band. - The device in the wireless multi-input multi-round (10) communication system includes: a controller, configured to derive a first transmit matrix between the receiving entities according to a first pilot received via a transmitting entity and a channel, and according to the estimated value of the channel response and the first And a second calibration error matrix derived - a second emission matrix, the ΜΙΜΟ channel response estimate is the estimated 胄 estimated, and derived from a second pilot received via the ΜΙΜΟ channel, the first calibration The differential Is vehicle includes an estimated value of an error in a first correction matrix for compensating for a response of a transmitting and receiving chain at the transmitting entity, and the second calibration error matrix includes a method for compensating for transmission and reception at the receiving entity Estimating an error in a second correction matrix of the response of the chain, and modifying the first and second calibration error matrices according to the first and second emission matrices; and a spatial processor operative to pass through the MlM The 〇 channel transmits the data payout number 'multiply it by the second correction matrix. 23. The apparatus of claim 22, wherein the first pilot is a guided pilot received via a plurality of eigenmodes of the channel, and wherein the second pilot is by the transmitting entity A pilot consisting of a plurality of pilot transmissions transmitted by a plurality of transmit antennas, wherein pilot transmissions from each of the 99590.doc 1361601 antennas are identifiable by the receiving entity. The apparatus of claim 22, wherein the controller is operative to modify the first and second calibration error matrices according to an adaptive procedure for adjusting the first and second calibration error matrices in an iterative manner to reduce The error between the first and first transmit matrices is small. The apparatus of claim 22, wherein the controller is further operative to decompose the MiMcmit response estimate to obtain the first eigenvector matrix, based on the ΜIΜ Ο channel response estimate and the And a second calibration error matrix, calculating a second eigenvector matrix of the ΜΙΜ0 channel,/ calculating the second emission matrix according to the second eigenvector matrix and the ΜΙΜ〇 channel response estimate. The apparatus of claim 22, wherein the controller is further operative to calculate a error matrix as a difference between the first and second emission matrices, and derive an element in the error matrix relative to the first a partial derivative of the selected element in the second calibration error matrix, 'the update vector is used to update the first and second more difference frames according to the partial derivative and the error matrix, and the moment is repeated to calculate the error material And deriving the partial derivatives, calculating the new vector', and updating the first and second calibration error matrices a plurality of times: until the update vector satisfies an end condition. 27. The apparatus of claim 22 wherein the controller is operative to calculate an error matrix as the difference between the first and second transmit matrices, 99590.doc 1361601 calculating a total error based on the error matrix, derived The total error is relative to the partial derivative of the selected element of the first and second calibration error matrices, and the first and second calibration error matrices are updated by using the partial derivatives, and the error matrix is repeated and the calculation is performed. The total error, deriving the partial derivatives, and updating the first and second calibration error matrices a plurality of times until the total error satisfies an end condition. 28. A gamut in a wireless multiple input multiple output (MIMO) communication system, comprising: a deriving means for receiving a first channel according to a channel between a transmitting entity and a receiving entity Pilot to derive a first transmit matrix; Deriving a component for deriving a second emission matrix according to a channel response estimate and the first and second calibration error matrices, wherein the μι μ channel response estimate is an estimated value of one of the channel responses, and Deriving from a second pilot received via the buffer channel, the first calibration error matrix includes an estimate of an error in a first correction matrix for compensating for a response of a transmit and receive chain at the transmit entity, and the a second calibration error matrix comprising: an estimate of an error in a second correction matrix for compensating for a response of the transmit and receive chains at the receiving entity; and a modifying component for modifying the first according to the first and second transmit matrices And a second calibration error matrix. 29. The apparatus of claim 28, wherein the first pilot system is a guided pilot frequency received via a plurality of eigenmodes of the channel, and wherein the second pilot system A chirp pilot formed by a plurality of pilot transmissions transmitted by a plurality of transmit antennas at the transmitting entity, wherein pilot transmissions from each of the transmit antennas are identifiable by the receiving entity. The apparatus of claim 28, further comprising: a decomposition component for decomposing the ΜΙΜ〇 channel response estimate to learn a first eigenvector matrix of the channel; the singular component' Calculating a second eigenvector matrix of the one of the chirp channels according to the chirp channel response estimate and the first and second calibration error matrices; and calculating means for using the second eigenvector matrix and the chirp channel The (four) two emission matrix is calculated in response to the estimated value. 31. The apparatus of claim 28, further comprising: a computing component for calculating an error matrix as a difference between the first and second emission matrices; and a deriving component 'for deriving an element in the error matrix relative to the a partial derivative of the selected element in the first and first weight error matrices; a computing component for calculating an update vector based on the partial derivative and the error matrix; updating the component, 纟 for updating using the update vector The first and second calibration error matrices; and a repeating component 'for repeating the calculating the error matrix, deriving the partial derivatives, counting the update vector, and updating the first and second calibration error matrices, Until the update vector satisfies an end condition. 99. The apparatus of claim 28, further comprising: a singular component for calculating an error matrix as a difference between the first 盥嗲 second emission matrix; an arithmetic component for using the error matrix Calculating a total error; deriving a component for deriving a partial derivative of the total error relative to a selected one of the first and second calibration error matrices; updating means for updating the first using the partial derivative And a first calibration error matrix; and a μ repeating component for repeating the calculating the error matrix, calculating the total error, deriving the partial derivatives, and updating the first and second calibration error matrices to a plurality of times The total error satisfies an end condition. 33. A method of calibrating a downlink and an uplink channel in a wireless multiple input multiple output (μιμ〇) communication system, comprising: a line chain under a μιμ〇 channel between a transmitting entity and a receiving entity And determining, by the path and uplink channel response estimates, a first calibration to obtain first and second correction matrices, the first correction matrix compensating for a response of the transmit and receive chains at the transmit entity, and the second correction The matrix is configured to compensate a response of the transmit and receive chains at the receiving entity; and perform a second calibration according to the first and second pilots received via the buffer channel to obtain first and second calibration error matrices, the first A calibration error matrix includes an estimate of the error in the first correction matrix, and the second calibration error matrix includes an estimate of the error in the second correction matrix. 34. The method of claim 33, further comprising: 99590.doc ^01601 using the second calibration error matrix to update the second correction matrix β 35. The method of claim 33, wherein the first pilot system Receiving a pilot pilot received by a plurality of eigenmodes of the milk assist channel, and wherein the first pilot is comprised of a plurality of pilot transmissions transmitted by a plurality of transmit antennas at the transmitting entity Pilots in which pilot transmissions from each transmit antenna are identifiable by the receiving entity. 36. The method of claim 33, wherein the performing - the second calibration comprises deriving a first-emission matrix from the first pilot, deriving a channel response estimate from the second pilot a first transmit matrix, and modifying the first and second calibration error matrices according to the first and second transmit matrices. The method of claim 36, wherein the first and second calibration error matrices are modified to reduce the first and second using an adaptive procedure for adjusting the first and second calibration error matrices in an iterative manner An error between the second transmit matrix. 38. A device in a wireless multiple input multiple output (MIMO) communication system that includes: a controller that operates between a transmitting entity and a receiving entity - ΜIΜ 0 channel downlink and uplink channel response estimates to perform a first order to obtain first and second correction matrices 'The first correction matrix is used to compensate for the transmit and receive chains at the transmitting entity Response, and the second correction matrix is used to compensate for the transmission and reception chain at the receiving entity 99590.doc 10 1361601 Performing a second calibration based on the first and second pilots received via the buffer channel to obtain first and second calibration error matrices, the first calibration error matrix including an estimate of the error in the first correction matrix, and The second calibration error matrix includes an estimate of the error in the second correction matrix; and two processors operative to multiply the data symbol by the second correction matrix before transmitting the data symbol via the M1M channel. 39. The apparatus of claim 38, wherein the controller is operative to derive a first transmit matrix based on the first pilot, and derive a second based on a channel response estimate obtained from the second pilot Transmitting a matrix, and modifying the first and second calibration error matrices according to the first and second emission matrices. 40. The apparatus of claim 39, wherein the controller is operative to change the first and second calibration errors using an adaptive procedure of the first and second calibration error matrices in a stacked mode a matrix to reduce the error between the first and second emission matrices. 41. Apparatus for a wireless multiple input multiple output (MIM(R)) communication system, comprising: means for performing a first calibration for receiving a channel between entities based on a transmitting entity The downlink and uplink channel response estimates are used to perform a second correction matrix, the first correction of the first calibration to obtain the first and the first matrix for compensating for the transmitting entity to send 99590.doc 8 -11 - 1361601 And a response of the receive chain, and the response two correction matrix of the transmit and receive chains of the first entity is used to compensate for the receive: and an execution component for performing the second calibration, for receiving the first received via the draw channel And performing: the second calibration to obtain the first and second calibration error matrices, the first calibration error matrix including an estimate of the error in the first correction matrix and the second calibration error matrix including the The second correction matrix is an estimate of the error of the towel. 42. The apparatus of claim 41, wherein the means for performing the second calibration comprises a derived component array for deriving a first emission moment deriving member based on the first pilot, the The second pilot obtains a channel response estimate to derive a second transmit matrix, and a modifying component for modifying the first and second calibration error matrices based on the first and second transmit matrices. 43· : 3⁄4 m 42 device 'which is used - adjusts the first and the second to adjust the adaptive program of the difference matrix to modify the first and second calibration error matrices, —时.^ ^ — 乂 Reduce the error between the first and second emission matrices 99590.doc -12-